generic-library/vpx
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity
f2ec452fbc
vpx/vp8/encoder/mbgraph.c
Adrian Grange 9daf3154db Superblock encoding order 
This is the first patch to add superblock (32x32) coding
order capabilities. It does not yet do any mode selection
at the SB level, that will follow in a further patch.

This patch encodes rows of SBs rather than
MBs, each SB contains 2x2 MBs.

Two intra prediction modes have been disabled since they
require reconstructed data for the above-right MB which
may not have been encoded yet (e.g. for the bottom right
MB in each SB).

Results on the one test clip I have tried (720p GIPS clip)
suggest that it is somewhere around 0.2dB worse than the
baseline version, so there may be bugs.

It has been tested with no experiments enabled and with
the following 3 experiments enabled:
  --enable-enhanced_interp
  --enable-high_precision_mv
  --enable-sixteenth_subpel_uv
in each case the decode buffer matches the recon buffer
(using "cmp" to compare the dumped/decoded frames).
Note: Testing these experiments individually created
errors.

Some problems were found with other experiments but it
is unclear what state these experiments are in:
  --enable-comp_intra_pred
  --enable-newentropy
  --enable-uvintra

This code has not been extensively tested yet, so there
is every likelihood that further bugs remain. I also
intend to do some code cleanup & refactoring in tandem
with the next patch that adds the 32x32 modes.

Change-Id: I1eba7f740a70b3510df58db53464535ef881b4d9
2012-04-11 10:40:57 +01:00

565 lines
18 KiB
C
Raw Blame History

/*
* Copyright (c) 2010 The WebM project authors. All Rights Reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#include <limits.h>
#include <vp8/encoder/encodeintra.h>
#include <vp8/encoder/rdopt.h>
#include <vp8/common/setupintrarecon.h>
#include <vp8/common/blockd.h>
#include <vp8/common/reconinter.h>
#include <vp8/common/systemdependent.h>
#include <vpx_mem/vpx_mem.h>
#include <vp8/encoder/segmentation.h>
static unsigned int do_16x16_motion_iteration
(
VP8_COMP *cpi,
int_mv *ref_mv,
int_mv *dst_mv
)
{
MACROBLOCK * const x = &cpi->mb;
MACROBLOCKD * const xd = &x->e_mbd;
BLOCK *b = &x->block[0];
BLOCKD *d = &xd->block[0];
vp8_variance_fn_ptr_t v_fn_ptr = cpi->fn_ptr[BLOCK_16X16];
unsigned int best_err;
int step_param, further_steps;
static int dummy_cost[2*mv_max+1];
int *mvcost[2] = { &dummy_cost[mv_max+1], &dummy_cost[mv_max+1] };
int *mvsadcost[2] = { &dummy_cost[mv_max+1], &dummy_cost[mv_max+1] };
#if CONFIG_HIGH_PRECISION_MV
static int dummy_cost_hp[2*mv_max_hp+1];
int *mvcost_hp[2] = { &dummy_cost_hp[mv_max_hp+1], &dummy_cost_hp[mv_max_hp+1] };
int *mvsadcost_hp[2] = { &dummy_cost_hp[mv_max_hp+1], &dummy_cost_hp[mv_max_hp+1] };
#endif
int col_min = (ref_mv->as_mv.col>>3) - MAX_FULL_PEL_VAL + ((ref_mv->as_mv.col & 7)?1:0);
int row_min = (ref_mv->as_mv.row>>3) - MAX_FULL_PEL_VAL + ((ref_mv->as_mv.row & 7)?1:0);
int col_max = (ref_mv->as_mv.col>>3) + MAX_FULL_PEL_VAL;
int row_max = (ref_mv->as_mv.row>>3) + MAX_FULL_PEL_VAL;
int tmp_col_min = x->mv_col_min;
int tmp_col_max = x->mv_col_max;
int tmp_row_min = x->mv_row_min;
int tmp_row_max = x->mv_row_max;
int_mv ref_full;
// Further step/diamond searches as necessary
if (cpi->Speed < 8)
{
step_param = cpi->sf.first_step + ((cpi->Speed > 5) ? 1 : 0);
further_steps = (cpi->sf.max_step_search_steps - 1) - step_param;
}
else
{
step_param = cpi->sf.first_step + 2;
further_steps = 0;
}
/* Get intersection of UMV window and valid MV window to reduce # of checks in diamond search. */
if (x->mv_col_min < col_min )
x->mv_col_min = col_min;
if (x->mv_col_max > col_max )
x->mv_col_max = col_max;
if (x->mv_row_min < row_min )
x->mv_row_min = row_min;
if (x->mv_row_max > row_max )
x->mv_row_max = row_max;
ref_full.as_mv.col = ref_mv->as_mv.col >> 3;
ref_full.as_mv.row = ref_mv->as_mv.row >> 3;
/*cpi->sf.search_method == HEX*/
best_err = vp8_hex_search(x, b, d,
&ref_full, dst_mv,
step_param,
x->errorperbit,
&v_fn_ptr,
#if CONFIG_HIGH_PRECISION_MV
xd->allow_high_precision_mv?mvsadcost_hp:mvsadcost, xd->allow_high_precision_mv?mvcost_hp:mvcost,
#else
mvsadcost, mvcost,
#endif
ref_mv);
// Try sub-pixel MC
//if (bestsme > error_thresh && bestsme < INT_MAX)
{
int distortion;
unsigned int sse;
best_err = cpi->find_fractional_mv_step(x, b, d,
dst_mv, ref_mv,
x->errorperbit, &v_fn_ptr,
#if CONFIG_HIGH_PRECISION_MV
xd->allow_high_precision_mv?mvcost_hp:mvcost,
#else
mvcost,
#endif
&distortion, &sse);
}
vp8_set_mbmode_and_mvs(x, NEWMV, dst_mv);
vp8_build_inter16x16_predictors_mby(xd);
//VARIANCE_INVOKE(&cpi->rtcd.variance, satd16x16)
best_err = VARIANCE_INVOKE(&cpi->rtcd.variance, sad16x16)
(xd->dst.y_buffer, xd->dst.y_stride,
xd->predictor, 16, best_err);
/* restore UMV window */
x->mv_col_min = tmp_col_min;
x->mv_col_max = tmp_col_max;
x->mv_row_min = tmp_row_min;
x->mv_row_max = tmp_row_max;
return best_err;
}
static int do_16x16_motion_search
(
VP8_COMP *cpi,
int_mv *ref_mv,
int_mv *dst_mv,
YV12_BUFFER_CONFIG *buf,
int buf_mb_y_offset,
YV12_BUFFER_CONFIG *ref,
int mb_y_offset
)
{
MACROBLOCK * const x = &cpi->mb;
MACROBLOCKD * const xd = &x->e_mbd;
unsigned int err, tmp_err;
int_mv tmp_mv;
int n;
for (n = 0; n < 16; n++) {
BLOCKD *d = &xd->block[n];
BLOCK *b = &x->block[n];
b->base_src = &buf->y_buffer;
b->src_stride = buf->y_stride;
b->src = buf->y_stride * (n & 12) + (n & 3) * 4 + buf_mb_y_offset;
d->base_pre = &ref->y_buffer;
d->pre_stride = ref->y_stride;
d->pre = ref->y_stride * (n & 12) + (n & 3) * 4 + mb_y_offset;
}
// Try zero MV first
// FIXME should really use something like near/nearest MV and/or MV prediction
xd->pre.y_buffer = ref->y_buffer + mb_y_offset;
xd->pre.y_stride = ref->y_stride;
//VARIANCE_INVOKE(&cpi->rtcd.variance, satd16x16)
err = VARIANCE_INVOKE(&cpi->rtcd.variance, sad16x16)
(ref->y_buffer + mb_y_offset,
ref->y_stride, xd->dst.y_buffer,
xd->dst.y_stride, INT_MAX);
dst_mv->as_int = 0;
// Test last reference frame using the previous best mv as the
// starting point (best reference) for the search
tmp_err = do_16x16_motion_iteration(cpi, ref_mv, &tmp_mv);
if (tmp_err < err)
{
err = tmp_err;
dst_mv->as_int = tmp_mv.as_int;
}
// If the current best reference mv is not centred on 0,0 then do a 0,0 based search as well
if (ref_mv->as_int)
{
int tmp_err;
int_mv zero_ref_mv, tmp_mv;
zero_ref_mv.as_int = 0;
tmp_err = do_16x16_motion_iteration(cpi, &zero_ref_mv, &tmp_mv);
if (tmp_err < err)
{
dst_mv->as_int = tmp_mv.as_int;
err = tmp_err;
}
}
return err;
}
static int do_16x16_zerozero_search
(
VP8_COMP *cpi,
int_mv *dst_mv,
YV12_BUFFER_CONFIG *buf,
int buf_mb_y_offset,
YV12_BUFFER_CONFIG *ref,
int mb_y_offset
)
{
MACROBLOCK * const x = &cpi->mb;
MACROBLOCKD * const xd = &x->e_mbd;
unsigned int err;
int n;
for (n = 0; n < 16; n++) {
BLOCKD *d = &xd->block[n];
BLOCK *b = &x->block[n];
b->base_src = &buf->y_buffer;
b->src_stride = buf->y_stride;
b->src = buf->y_stride * (n & 12) + (n & 3) * 4 + buf_mb_y_offset;
d->base_pre = &ref->y_buffer;
d->pre_stride = ref->y_stride;
d->pre = ref->y_stride * (n & 12) + (n & 3) * 4 + mb_y_offset;
}
// Try zero MV first
// FIXME should really use something like near/nearest MV and/or MV prediction
xd->pre.y_buffer = ref->y_buffer + mb_y_offset;
xd->pre.y_stride = ref->y_stride;
//VARIANCE_INVOKE(&cpi->rtcd.variance, satd16x16)
err = VARIANCE_INVOKE(&cpi->rtcd.variance, sad16x16)
(ref->y_buffer + mb_y_offset,
ref->y_stride, xd->dst.y_buffer,
xd->dst.y_stride, INT_MAX);
dst_mv->as_int = 0;
return err;
}
static int find_best_16x16_intra
(
VP8_COMP *cpi,
YV12_BUFFER_CONFIG *buf,
int mb_y_offset,
MB_PREDICTION_MODE *pbest_mode
)
{
MACROBLOCK * const x = &cpi->mb;
MACROBLOCKD * const xd = &x->e_mbd;
MB_PREDICTION_MODE best_mode = -1, mode;
int best_err = INT_MAX;
// calculate SATD for each intra prediction mode;
// we're intentionally not doing 4x4, we just want a rough estimate
for (mode = DC_PRED; mode <= TM_PRED; mode++)
{
unsigned int err;
xd->mode_info_context->mbmi.mode = mode;
RECON_INVOKE(&cpi->rtcd.common->recon, build_intra_predictors_mby)(xd);
//VARIANCE_INVOKE(&cpi->rtcd.variance, satd16x16)
err = VARIANCE_INVOKE(&cpi->rtcd.variance, sad16x16)
(xd->predictor, 16,
buf->y_buffer + mb_y_offset,
buf->y_stride, err);
// find best
if (err < best_err)
{
best_err = err;
best_mode = mode;
}
}
if (pbest_mode)
*pbest_mode = best_mode;
return best_err;
}
static void update_mbgraph_mb_stats
(
VP8_COMP *cpi,
MBGRAPH_MB_STATS *stats,
YV12_BUFFER_CONFIG *buf,
int mb_y_offset,
YV12_BUFFER_CONFIG *golden_ref,
int_mv *prev_golden_ref_mv,
int gld_y_offset,
YV12_BUFFER_CONFIG *alt_ref,
int_mv *prev_alt_ref_mv,
int arf_y_offset
)
{
MACROBLOCK * const x = &cpi->mb;
MACROBLOCKD * const xd = &x->e_mbd;
int intra_error;
// FIXME in practice we're completely ignoring chroma here
xd->dst.y_buffer = buf->y_buffer + mb_y_offset;
// do intra 16x16 prediction
intra_error = find_best_16x16_intra(cpi, buf, mb_y_offset, &stats->ref[INTRA_FRAME].m.mode);
if (intra_error <= 0)
intra_error = 1;
stats->ref[INTRA_FRAME].err = intra_error;
// Golden frame MV search, if it exists and is different than last frame
if (golden_ref)
{
int g_motion_error = do_16x16_motion_search(cpi, prev_golden_ref_mv,
&stats->ref[GOLDEN_FRAME].m.mv,
buf, mb_y_offset,
golden_ref, gld_y_offset);
stats->ref[GOLDEN_FRAME].err = g_motion_error;
}
else
{
stats->ref[GOLDEN_FRAME].err = INT_MAX;
stats->ref[GOLDEN_FRAME].m.mv.as_int = 0;
}
// Alt-ref frame MV search, if it exists and is different than last/golden frame
if (alt_ref)
{
//int a_motion_error = do_16x16_motion_search(cpi, prev_alt_ref_mv,
// &stats->ref[ALTREF_FRAME].m.mv,
// buf, mb_y_offset,
// alt_ref, arf_y_offset);
int a_motion_error =
do_16x16_zerozero_search( cpi,
&stats->ref[ALTREF_FRAME].m.mv,
buf, mb_y_offset,
alt_ref, arf_y_offset);
stats->ref[ALTREF_FRAME].err = a_motion_error;
}
else
{
stats->ref[ALTREF_FRAME].err = INT_MAX;
stats->ref[ALTREF_FRAME].m.mv.as_int = 0;
}
}
static void update_mbgraph_frame_stats
(
VP8_COMP *cpi,
MBGRAPH_FRAME_STATS *stats,
YV12_BUFFER_CONFIG *buf,
YV12_BUFFER_CONFIG *golden_ref,
YV12_BUFFER_CONFIG *alt_ref
)
{
MACROBLOCK * const x = &cpi->mb;
VP8_COMMON * const cm = &cpi->common;
MACROBLOCKD * const xd = &x->e_mbd;
int mb_col, mb_row, offset = 0;
int mb_y_offset = 0, arf_y_offset = 0, gld_y_offset = 0;
int_mv arf_top_mv, gld_top_mv;
MODE_INFO mi_local;
// Set up limit values for motion vectors to prevent them extending outside the UMV borders
arf_top_mv.as_int = 0;
gld_top_mv.as_int = 0;
x->mv_row_min = -(VP8BORDERINPIXELS - 16 - INTERP_EXTEND);
x->mv_row_max = (cm->mb_rows - 1) * 16 + VP8BORDERINPIXELS - 16 - INTERP_EXTEND;
xd->up_available = 0;
xd->dst.y_stride = buf->y_stride;
xd->pre.y_stride = buf->y_stride;
xd->dst.uv_stride = buf->uv_stride;
xd->mode_info_context = &mi_local;
for (mb_row = 0; mb_row < cm->mb_rows; mb_row++)
{
int_mv arf_left_mv, gld_left_mv;
int mb_y_in_offset = mb_y_offset;
int arf_y_in_offset = arf_y_offset;
int gld_y_in_offset = gld_y_offset;
// Set up limit values for motion vectors to prevent them extending outside the UMV borders
arf_left_mv.as_int = arf_top_mv.as_int;
gld_left_mv.as_int = gld_top_mv.as_int;
x->mv_col_min = -(VP8BORDERINPIXELS - 16 - INTERP_EXTEND);
x->mv_col_max = (cm->mb_cols - 1) * 16 + VP8BORDERINPIXELS - 16 - INTERP_EXTEND;
xd->left_available = 0;
for (mb_col = 0; mb_col < cm->mb_cols; mb_col++)
{
MBGRAPH_MB_STATS *mb_stats = &stats->mb_stats[offset + mb_col];
update_mbgraph_mb_stats(cpi, mb_stats, buf, mb_y_in_offset,
golden_ref, &gld_left_mv, gld_y_in_offset,
alt_ref, &arf_left_mv, arf_y_in_offset);
arf_left_mv.as_int = mb_stats->ref[ALTREF_FRAME].m.mv.as_int;
gld_left_mv.as_int = mb_stats->ref[GOLDEN_FRAME].m.mv.as_int;
if (mb_col == 0)
{
arf_top_mv.as_int = arf_left_mv.as_int;
gld_top_mv.as_int = gld_left_mv.as_int;
}
xd->left_available = 1;
mb_y_in_offset += 16;
gld_y_in_offset += 16;
arf_y_in_offset += 16;
x->mv_col_min -= 16;
x->mv_col_max -= 16;
}
xd->up_available = 1;
mb_y_offset += buf->y_stride * 16;
gld_y_offset += golden_ref->y_stride * 16;
if (alt_ref)
arf_y_offset += alt_ref->y_stride * 16;
x->mv_row_min -= 16;
x->mv_row_max -= 16;
offset += cm->mb_cols;
}
}
// Test for small magnitude (<= 1 pel mvs)
int small_mv( MV mv )
{
if ( (abs( (int)mv.col ) > 2) || (abs( (int)mv.row ) > 2) )
return FALSE;
else
return TRUE;
}
//void separate_arf_mbs_byzz
void separate_arf_mbs
(
VP8_COMP *cpi
)
{
VP8_COMMON * const cm = &cpi->common;
int mb_col, mb_row, offset, i;
int ncnt[4];
int n_frames = cpi->mbgraph_n_frames;
int * arf_not_zz;
CHECK_MEM_ERROR(arf_not_zz,
vpx_calloc(cm->mb_rows * cm->mb_cols * sizeof(*arf_not_zz), 1));
vpx_memset(arf_not_zz, 0, sizeof(arf_not_zz));
// We are not interested in results beyond the alt ref itself.
if ( n_frames > cpi->frames_till_gf_update_due )
n_frames = cpi->frames_till_gf_update_due;
// defer cost to reference frames
for (i = n_frames - 1; i >= 0; i--)
{
MBGRAPH_FRAME_STATS *frame_stats = &cpi->mbgraph_stats[i];
for (offset = 0, mb_row = 0; mb_row < cm->mb_rows;
offset += cm->mb_cols, mb_row++)
{
for (mb_col = 0; mb_col < cm->mb_cols; mb_col++)
{
MBGRAPH_MB_STATS *mb_stats =
&frame_stats->mb_stats[offset + mb_col];
int altref_err = mb_stats->ref[ALTREF_FRAME].err;
int intra_err =
mb_stats->ref[INTRA_FRAME ].err + 250;
int golden_err =
mb_stats->ref[GOLDEN_FRAME].err + 250;
// Test for altref vs intra and gf and that its mv was 0,0.
if ( (altref_err > 1000) ||
(altref_err > intra_err) ||
(altref_err > golden_err) )
{
arf_not_zz[offset + mb_col]++;
}
}
}
}
vpx_memset(ncnt, 0, sizeof(ncnt));
for (offset = 0, mb_row = 0; mb_row < cm->mb_rows;
offset += cm->mb_cols, mb_row++)
{
for (mb_col = 0; mb_col < cm->mb_cols; mb_col++)
{
// If any of the blocks in the sequence failed then the MB
// goes in segment 0
if ( arf_not_zz[offset + mb_col] )
{
ncnt[0]++;
cpi->segmentation_map[offset + mb_col] = 0;
}
else
{
ncnt[1]++;
cpi->segmentation_map[offset + mb_col] = 1;
}
}
}
// Only bother with segmentation if over 10% of the MBs in static segment
//if ( ncnt[1] && (ncnt[0] / ncnt[1] < 10) )
if ( 1 )
{
// Note % of blocks that are marked as static
if ( cm->MBs )
cpi->static_mb_pct = (ncnt[1] * 100) / cm->MBs;
// This error case should not be reachable as this function should
// never be called with the common data structure unititialized.
else
cpi->static_mb_pct = 0;
vp8_enable_segmentation((VP8_PTR) cpi);
}
else
{
cpi->static_mb_pct = 0;
vp8_disable_segmentation((VP8_PTR) cpi);
}
// Free localy allocated storage
vpx_free(arf_not_zz);
}
void vp8_update_mbgraph_stats
(
VP8_COMP *cpi
)
{
VP8_COMMON * const cm = &cpi->common;
int i, n_frames = vp8_lookahead_depth(cpi->lookahead);
YV12_BUFFER_CONFIG *golden_ref = &cm->yv12_fb[cm->gld_fb_idx];
// we need to look ahead beyond where the ARF transitions into
// being a GF - so exit if we don't look ahead beyond that
if (n_frames <= cpi->frames_till_gf_update_due)
return;
if (n_frames > MAX_LAG_BUFFERS)
n_frames = MAX_LAG_BUFFERS;
cpi->mbgraph_n_frames = n_frames;
for (i = 0; i < n_frames; i++)
{
MBGRAPH_FRAME_STATS *frame_stats = &cpi->mbgraph_stats[i];
vpx_memset(frame_stats->mb_stats, 0,
cm->mb_rows * cm->mb_cols * sizeof(*cpi->mbgraph_stats[i].mb_stats));
}
// do motion search to find contribution of each reference to data
// later on in this GF group
// FIXME really, the GF/last MC search should be done forward, and
// the ARF MC search backwards, to get optimal results for MV caching
for (i = 0; i < n_frames; i++)
{
MBGRAPH_FRAME_STATS *frame_stats = &cpi->mbgraph_stats[i];
struct lookahead_entry *q_cur =
vp8_lookahead_peek(cpi->lookahead, i);
assert(q_cur != NULL);
update_mbgraph_frame_stats(cpi, frame_stats, &q_cur->img,
golden_ref, cpi->Source);
}
vp8_clear_system_state(); //__asm emms;
separate_arf_mbs(cpi);
}
Reference in New Issue View Git Blame Copy Permalink